US20040133709A1

US20040133709A1 - Method and system for personalized I/O device initialization

Info

Publication number: US20040133709A1
Application number: US10/614,965
Authority: US
Inventors: Janko Boehm; Herwig Elfering; Thomas Hess; Daniel Metz
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2002-07-09
Filing date: 2003-07-08
Publication date: 2004-07-08
Also published as: US20060015657A1; US7725612B2

Abstract

A method for controlling a plurality of I/O devices being attached to a microprocessor by a special number and type of interfaces is provided. A generic configurable chip is connected to the I/O space of said microprocessor, said generic chip comprising a switch matrix being adapted to assign said special number and type of interfaces to each specific device when initializing said device.

Description

BACKGROUND OF THE INVENTION

The present invention relates, in general, to accessing I/O functions. More specifically, the invention relates to accessing a great variance of I/O devices in a control system. Still more specifically, the present invention deals with accessing such I/O functions in embedded control environments.

In embedded control applications it is not always possible to meet all wanted requirements with standard I/O chips. Very often, it is tried to fulfill these requirements with a custom design. This design then will end in realization of a special designed chip, based, e.g., on ASICs. However, typically the I/O connectivity of these chips and not the available space is limiting the integration of more functionality on the same chip.

Though I/O devices may have a high similarity according to their requirements with regard to the I/O protocols supported (e.g., IC bus (I ²C), Universal Asynchronous Receiver/Transmitter (UART), General Purpose I/O (GPI/O) and the like), the number of I/O ports is different on each specific device.

In many applications, a certain number and type of I/O devices must be controlled (it is of no importance whether these devices are cards in a computer system, switches or hub-devices in a network or switching units in a luggage or parcel sorting system or similar systems). Typically, this controlling task is done by means of a microcontroller. However, in most cases, the industry does not offer out of the shelf a respective controller with enough or exactly the right I/O interfaces on the core chip. Accordingly, a chip (most probably an ASIC) will have to be developed that can be connected to the I/O space (memory mapped I/O, PCI, CAN-bus, Ethernet, etc.) of the controller as an extender chip. As already mentioned above, this chip would most probably be a custom design and is expanding the number and the type of the controller interfaces in order to exactly offer the right number and type of interfaces as required by the devices to perform their specific tasks. However, even if all of these devices would need the same sort or type of I/O interfaces (such as GPI/O, UART, IIC, and the like), the quantity of the interfaces may be different for each of the different device types to be controlled. The traditional methods to solve this problem would be to a) develop one individual chip (ASIC) for each device type or, b) develop one single chip (ASIC) providing a superset (in number and type) of all the interfaces needed for all the different device types.

The disadvantage of solution a) is obvious as it is clear that developing and producing n (assumed n being the number of different device types that shall be controlled) different chips (ASICs) is much more expensive then doing all this just once for one chip being produced in bigger volumes.

However, using solution b) would not only maximize the number of transistors on the chip, but would at the same time maximize the number of pins. Since industry is able to double the transistor density on a chip about every 18 months, the number of I/O pins is more and more determining the size and at the same time the costs (for each individual chip as well as for the space in the cards) of a chip.

BRIEF SUMMARY OF THE INVENTION

It is therefore an object of the present invention to provide one single chip providing a superset of all the interfaces needed for all the different device types, this chip not being bigger than any individual chip for each device type.

It is a further object of the present invention to provide a method to perform an individual initialization procedure at the time the device gets attached to a system.

It is still a further object of the invention to do this initialization when new hardware is attached to a running system, in computer technique known as hot-plugging.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING

These and other objects will be apparent to one skilled in the art from the following detailed description of the invention taken in conjunction with the accompanying drawings in which: [0010]
FIG. 1 is a flow-chart indicating the initialization sequence according to the invention; [0011]
FIG. 2 schematically shows a first embodiment of a control network according to the invention; [0012]
FIG. 3 schematically shows a second embodiment of a control network according to the invention; and [0013]
FIG. 4 schematically shows how the invention is employed on a chip.[0014]

DETAILED DESCRIPTION OF THE INVENTION

With the present invention, a flexible way is provided to assign I/O connectivity to different function units on a chip, depending on the functionality needed for the specific field of operation the chip will be employed for. This requires a special initialization step to personalize the hardware according to the special requirements of the specific device during the device driver initialization. To do so allows to build a more generic hardware for a certain variance of I/O devices which could be produced with lower costs in a higher volume. [0015]
The following table gives an example for three different devices needing various numbers of the three interface types GPI/O, UART and IIC. It has to be mentioned that the present invention is not restricted to a special number or type of devices or interfaces but can be used with any number and type of devices and interfaces. [0016]

Device Device Device

Type A Type B Type C Max

I/O # of # of # of # of # of # of # of

Type interfaces pins interfaces pins interfaces pins pins

GPI/O 48 48 20 20 60 60 60

IIC 10 20 14 28 8 16 28

UART 30 60 40 80 24 48 80

Sum 128 128 124 168
In case one would develop one individual chip (ASIC) for each device type, three different chips would be needed, each of them not bigger than 128 pins. It has to be mentioned that this is only an academic example where just the functional pins are taken into account. The power supply pins, e.g., are not counted in this example. However, this has no influence on the subject-matter of the invention. [0017]
In case one single chip (ASIC) providing a superset (in number and type) of all the interfaces needed for all the different device types is to be developed, this universal chip would need at least 168 (this being the sum of the maximum number of pins for each of the three devices given in the table), if not 196 pins (depending on the granularity of the chips. This, however, would enlarge the chip size by at least 31%. [0018]
The present invention solves this problem by providing a universal, generic chip which reduces the number of pins to the highest number of pins of the biggest individual chip, i.e., 128 in the above example. These 128 pins are the maximum of the three individual chips mentioned above. The reduction is achieved by a switch matrix that is assigning the pins according to the needs of the specific device. The assignment of this switch matrix can be done either [0019]
a) in hardware by the use of ID-bits on the card. These ID-bits would be implemented a select lines preselecting the functions of the chip required in the specific field of operation; or [0020]
b) by software initialization. [0021]
Method a) is inflexible. Depending on the source of identification, e.g., ID-bits which are defined due to specific patterns on default GP-I/O lines, the pins are assigned by the switch matrix to the pin-out appropriate for the device according to the ID. The disadvantage is that, at the point in time when the chip is designed, all IDs and their exact pin assignments will have to be known. The software method b) is more flexible and will therefore be described in the following in more detail. [0022]
The main idea is that the controller prior to any control task is setting up the chip (ASIC), according to the device that has to be controlled. The chip may consist either of an active (e.g., a microprocessor design MACROS together with the custom design MACRO on one chip) or an inactive (ASIC, just the custom design on the chip) element. After this set up the chip can be used by the controller software for the intrinsic controlling task in the same way as any traditional device. To do this setup, the software has to identify the chip or the device before the switch matrix can be initialized accordingly. It is proposed that the identification can be done, e.g., by reading the ID-bits, using the default settings of the switch matrix (cf. FIG. 4). To support this, the switch matrix control unit will always force the switch matrix in this state after initial power-on of the chip. The requirement for an identification is not a special additional requirement of the present invention. Whenever the controller controls a number of different devices, a device identification is needed. This is independent for the proposed solution. Therefore, the requirement is not evoked from the present invention. [0023]
In the above mentioned example, the identification would be done by data stored in an SEEPROM. The SEEPROM would be connected to the chip by an IIC bus or in any other possible way, e.g., via UART, GPI/O, etc. In this case the default connection of the matrix would always—this especially also means before any initialization—support this default connection to the IIC engine in the chip. The identification could of course also be done in other ways, for example by reading some ID-bits via the GPI/O engine, or by using the UART engine for communication to another chip on the card to get the unambiguous identifier of the device. In both latter cases, of course, the GPI/O respectively the UART engine must be connected to fixed output pins by the default matrix. In any case the initialization sequence would look like in FIG. 1. In the first step ([0024] 1) the device driver would be initialized so that the device can be accessed at all. Subsequently, the default connection on the device would be initialized (2). In the above example this would be the IIC device to access the identification information. In the next step (3) this information would be read. The code would use this information to set the matrix according to requirements of the identified device. This would be done before in the next step (4) the code would initialize the now known hardware of the entire device and the device driver. If all steps were completed successfully the code now can start using the full functionality of this device (5). If any of the steps 1 to 5 fail, the code would execute (7), report the error and the device would stay uninitialized and can not be used for further work.
The purpose of the invention is the requirement to control a higher number of different card types or the like. These cards may be installed in a computer frame and, according to their type, have different requirements regarding their I/O connectivity, depending on the task that has to be served on the specific card. For all these cards the generic chip is more or less the entrance of control for the embedded controller. For this purpose all these chips are connected with the controller by a proprietary serial interface. [0025]
This type of interface and protocol could be replaced by any communication infrastructure (like IIC, UART, Ethernet, USB, PCI-Bus, Firewire, or others), and protocol (like TCP/IP, PPP, etc.), or it could be even attached directly to the memory bus of the controller. Whenever special hardware has to be designed, because the I/O capacity of standard controllers is not sufficient, the present invention is applicable for the control task. [0026]
Given the present invention, in these cases one generic chip can be used to serve the requirements of all the different devices. Other operational fields could be the control of an assembly belt or for instance a stack of different hubs, switches, or other network devices that shall be controlled by one central point of control. A first way to build such a control network is shown in FIG. 2. Here a [0027] centralized microprocessor 2 is talking to the distributed devices (not shown) each having the generic chip 6 according to the invention via a chip protocol 4 like IIC.
The other way would be an arrangement as depicted in FIG. 3, where a [0028] main controller 8 is used. This controller 8 is connected to smaller autarchic distributed controllers 10 (auxiliary controller) which talk to each others via Ethernet 12 or any other Network interface. In this case the generic chip according to the invention would be something like an extender chip to the auxiliary controllers to generate the required I/O fan out.
FIG. 4 schematically depicts how the switch matrix is employed on the chip. The matrix itself could be implemented in any state of the art implementation as, e.g., known from crossbar switches used in telecommunication hardware. [0029]
On top of FIG. 4, a [0030] unit 14 of the chip 6 is depicted, which is building the interface, that is connecting the chip and the controller. This interface can be any type of interface applicable for inter chip connection. In case of a passive implementation of the chip no processing unit is part of interface unit. This means, the protocol connecting the chip with an outer controller needs to be applicable for passive chips (protocols like: IIC and the like. In case of an active implementation of the chip the interface unit would comprise a processing unit. In this case for the communication to the main controller and communication protocol applicable for the communication between processors (e.g. UART, ethernet, and the like) can be used to communicate to the chip. An on-chip bus 16 is used to connect the interface unit 14 with a control unit 18 of the switch matrix and a certain number of working units 20. This working units 20 are used to serve all the different types of interfaces needed of the controlled card (e.g. IIC, UART, GPI/O, JTAG, and the like). The output lines of these working units are connected to the switch matrix 22. This switch matrix is the main hardware requirement for the present invention. It allows to connect the n output lines of the working units 20 to the m pins of the chip via I/O connections 24, where n is (much) bigger than m. The switch matrix control unit allows to assign up to m of the n lines to the m pins. This assignment has to be done in the initialization step. By default, without any special assignment some lines j (j=n−1) of a working unit are connected to the j pins (j=m−k) of a default I/O connection 26. These j pins are connected to, e.g., ID-Bits or any other source of identification 28 (e.g., SEEPROM via IIC bus according to the example). After the initialization the chip can be used to perform any control task for a specific card not needing more then m pins.
The following is a short summary of the different steps of the method according to the present invention. [0031]
In the first step the device driver is initialized by the software. This is the start point of the following initialization process. As already mentioned, the software needs a default access path to get the information to identify the device. In the special case this default path is the path to the IIC unit. In this step the software resets and initializes this default path. Via the known and predefined default path, the software reads the identification information. In the present case this information will be read via IIC from the SEEPROM. The software keeps a list of known and supported chips according to the read ID. Depending on the ID, the software then sets the switch matrix. After initialization and configuration of the switch matrix the universal chip works according to the ID found in the SEEPROM content. This means the pins are set up and from now on the software can work on the universal chip like it would do with a specific type of chip. [0032]
Depending on the matrix configuration (in our example for device A, B or C) the units of the chip have to be initialized in a specific way. According to the already known ID, the controller can execute the appropriate initialization procedures of the chip units to be able to handle the specific device. In case of success of all earlier steps the chip can now be used by higher software layers. [0033]
While the preferred embodiment of the invention has been illustrated and described herein, it is to be understood that the invention is not limited to the precise construction herein disclosed, and the right is reserved to all changes and modifications coming within the scope of the invention as defined in the appended claims. [0034]

Claims

What is claimed is:

1. A method for controlling a plurality of I/O devices being attached to a microprocessor by a special number and type of interfaces comprising:

connecting a configurable chip to the I/O space of said microprocessor, said configurable chip having a switch matrix;

initializing said I/O devices; and

assigning to said switch matrix, said special number and type of interfaces to each I/O device during said initialization.

2. The method according to claim 1, wherein said configurable chip is an active element.

3. The method according to claim 2, wherein said configurable chip is a controller.

4. The method according to claim 1, wherein said configurable chip is a passive element.

5. The method according to claim 4, wherein said configurable chip is an ASIC.

6. The method according to claim 1, wherein said I/O devices support different I/O protocols.

7. The method according to claim 6, wherein said different I/O protocols comprise IC bus (I²C), Universal Asynchronous Receiver/Transmitter (UART), General Purpose I/O (GPI/O) protocols.

8. The method according to claim 1, wherein said switch matrix is adapted to assign I/O pins according to the needs of each I/O device.

9. The method according to claim 1 further comprising switching in hardware using ID bits on each I/O device.

10. The method according to claim 1, wherein each I/O device contains configuration data.

11. The method according to claim 10, wherein said configuration data includes an identifier for each of said I/O devices.

12. The method according to claim 10, wherein said configuration data includes initialization data.

13. The method according to claim 12, wherein said configuration data is stored in an SEEPROM on said I/O device.

14. The method according to claim 11, wherein said identifier is transmitted to said configurable chip.

15. The method according to any one of claims 9 wherein said switching is done by software initialization.

16. The method according to claim 15, wherein said microprocessor sets up said chip prior to any control task with regard to the device to be controlled.

17. The method according to claim 16, wherein said set up is done by the software identifying each specific device.

18. The method according to claim 16, wherein initialization of said specific device is done after said set up.

19. The method according to claim 1 wherein said initialization is started each time a new I/O device is attached to said microprocessor.

20. The method according to claim 19, wherein said new I/O device is hot plugged to a system.

21. The method according to claim 20, wherein hot plugging is indicated by an interrupt.

22. A computer system comprising:

a microprocessor for controlling I/O devices;

a plurality of I/O devices attached to said microprocessor by a special number and type of interfaces;

a configurable chip connected to the I/O space of said microprocessor, said configurable chip having a switch matrix;

an initializing program initializing one of said I/O devices; and

said initializing program assigning to said switch matrix, said special number and type of interfaces to each I/O device during said initialization.

23. The system according to claim 22, wherein said configurable chip is an active element.

24. The system according to claim 23, wherein said configurable chip is a controller.

25. The system according to claim 22, wherein said configurable chip is a passive element.

26. The system according to claim 25, wherein said configurable chip is an ASIC.

27. The system according to claim 22, wherein said I/O devices support different I/O protocols.

28. The system according to claim 27, wherein said different I/O protocols comprise IC bus (I²C), Universal Asynchronous Receiver/Transmitter (UART), General Purpose I/O (GPI/O) protocols.

29. The system according to claim 22, wherein said switch matrix is adapted to assign I/O pins according to the needs of each I/O device.

30. The system according to claim 22 further comprising switching in hardware using ID bits on each I/O device.

31. The system according to claim 22, wherein each I/O device contains configuration data.

32. The system according to claim 31, wherein said configuration data includes an identifier for each of said I/O devices.

33. The system according to claim 31, wherein said configuration data includes initialization data.

34. The system according to claim 33, wherein said configuration data is stored in an SEEPROM on said I/O device.

35. The system according to claim 32, wherein said identifier is transmitted to said configurable chip.

36. The system according to any one of claims 30 wherein said switching is done by software initialization.

37. The system according to claim 36, wherein said microprocessor sets up said chip prior to any control task with regard to the device to be controlled.

38. The system according to claim 37, wherein said set up is done by the software identifying each specific I/O device.

39. The system according to claim 37, wherein initialization of said specific device is done after said set up.

40. The system according to claim 22, wherein said initialization is started each time a new I/O device is attached to said microprocessor.

41. The system according to claim 40, wherein said new I/O device is hot plugged to a system.

42. The system according to claim 41, wherein hot plugging is indicated by an interrupt.

43. A program product for controlling a plurality of I/O devices being attached to a microprocessor by a special number and type of interface and having a generic configurable chip connected to the I/O space of the microprocessor, the configurable chip having a switch matrix, said program product comprising:

a computer readable medium having recorded thereon computer readable program code performing the method comprising:

initializing said I/O devices; and

44. The program product according to claim 43, wherein said I/O devices support different I/O protocols.

45. The program product according to claim 44, wherein said different I/O protocols comprise IC bus (I²C), Universal Asynchronous Receiver/Transmitter (UART), General Purpose I/O (GPI/O) protocols.

46. The program product according to claim 43 further comprising switching in hardware using ID bits on each I/O device.

47. The program product according to claim 43, wherein each I/O device contains configuration data.

48. The program product according to claim 47, wherein said configuration data includes an identifier for each of said I/O devices.

49. The program product according to claim 47, wherein said configuration data includes initialization data.

50. The program product according to claim 49, wherein said configuration data is stored in an SEEPROM on said I/O device.

51. The program product according to claim 48, wherein said identifier is transmitted to said configurable chip.

52. The program product according to any one of claims 46 wherein said switching is done by software initialization.

53. The program product according to claim 52, wherein said microprocessor sets up said chip prior to any control task with regard to the device to be controlled.

54. The program product according to claim 53, wherein said set up is done by the software identifying each specific device.

54. The program product according to claim 53, wherein initialization of said specific device is done after said set up.

55. The program product according to claim 43 wherein said initialization is started each time a new I/O device is attached to said microprocessor.

56. The program product according to claim 55, wherein said new I/O device is hot plugged to a system.

57. The program product according to claim 56, wherein hot plugging is indicated by an interrupt.